A Distant Supervision Method for Product Aspect Extraction from Customer Reviews

In this paper, we describe a distant supervision approach for the task of detecting product aspect mentions in customer reviews (e.g., in hotel reviews, we want to associate the aspect ``sleep quality" to a sentence such as "We both slept like rocks."). Detecting such aspects represents an important subtask of aspect-oriented review mining systems, which aim at automatically generating structured summaries of customer opinions. The main advantage of the proposed method is that it allows for the high accuracy of a supervised approach and at the same time avoids the costs of manually labeling a training set. We show how to exploit the inherent structure of customer reviews to automatically gather large amounts of labeled data. Our experimental results show that the method achieves a performance as good as a traditional, fully supervised approach.